Quality Assessment of modelled protein structure using Back-propagation and Radial Basis Function algorithm

• Published in: International Journal of Scientific Research and Management (IJSRM) Vol. 5; no. 7; pp. 6019 - 6033
• Main Author: Dr. Baljit Singh Khehra, Er. Amanpreet Kaur
• Format: Journal Article
• Language: English
• Published: 04.07.2017
• ISSN: 2321-3418; 2321-3418
• DOI: 10.18535/ijsrm/v5i7.27

Protein structure prediction (PSP) is the most important and challenging problem in bioinformatics today. This is due to the fact that the biological function of the protein is determined by its structure. While there is a gap between the number of known protein structures and the number of known protein sequences, protein structure prediction aims at reducing this structure –sequence gap. Protein structure can be experimentally determined using either X-ray crystallography or Nuclear Magnetic Resonance (NMR). However, these empirical techniques are very time consuming. So, various machine learning approaches have been developed for protein structure prediction like HMM, SVM and NN. In this paper, general introductory background to the area is discussed and two approaches of neural network i.e backpropagation and radial basis function are used for the prediction of protein tertiary structure. The aim of the study is to observe performance and applicability of these two neural network approaches on the same problem. More specifically, feed-forward artificial neural networks are trained with backpropagation neural network and radial basis function neural networks. These algorithms are used for the classification of protein data set, trained with the same input parameters and output data so that they can be compared. The advantages and disadvantages, in terms of the quality of the results, computational cost and time are identified. An algorithm for the selection of the spread constant is applied and tests are performed for the determination of the neural network with the best performance. These approaches depends on the chemical and physical properties of the constituent amino acids. Not all neural network algorithms have the same performance, so we represent the general success keys for any such algorithm. The data set used in the study is available as supplement at http://bit.ly/RF-PCP-DataSets.